I had 16/128GB version before which unfortunately has damaged debug console. I purchased a replacement recently.
The problem is that sometimes, it just freezes without a kernel panic or kdb prompt (I have both configured). I configured my kernel to drop to kdb on ttyS0 (UART) and verified that it works when I hit Akt-SysRq-C
It happened on both boards, so I think this is not a board problem. I even tested memory both from uboot and from Linux, booting with memtest parameter, and using memtester program found no problems with RAM.
I enabled a watchdog in dts files and it works. It will reset system if no userspace kick will happen in 2mins. This was also tested.
So, this night I saw my system was reset by watchdog.
My peripherals on this system:
- 2TB Kingston SKC3000D2048G SSD sitting in m.2 slot
- 1TB Sandisk microsd card (slow cheap one)
- BPI-R4 PWM fan
- ath9k Wi-Fi card for experiments, disabled for most of the times
The main Wi-Fi chip is not used. Board is connected to a second ethernet port. The speed of this port matches 1GBit.
The device is powered by 12VDC 36W EATON DC UPS, which shares the 12V bus with a 5G router. The avg. power draw, measured by 100V 20A capable DC wattmeter is around 7W, at high loads it reaches 10W~11W.
Highest uptime was 41 days in this configuration, then I took it down for full offline backup.
Could it be caused by sudden power spikes generated by SSD? Shall I downgrade it to a less powerhungry one? So far I also have 1TB SSD in my drawer but I would be very reluctant to moving large amounts of data offline from the device.
The system storage is fully encrypted by AES cipher, R/W speed to encrypted SSD are around 80MB/s.
Soon another same unit will be shipped to me, I will setup a day long stress-ng test of all of its subsystems on a completely empty system booted off sdcard.